Advancing Emotion Recognition through Innovative EEG Feature Extraction Techniques using S-Transform

The proposed study aimed to discover the development and optimization of an innovative EEG feature extraction technique, namely the S-Transform, for emotion recognition. In this regard, a data collection was conducted from 150 participants who were stimulated by various emotional stimuli during the experiment. EEG signals were recorded and pre-processed to extract features. Applying the S-Transform towards extracting time-frequency features of EEG signals, machine learning classifiers, such as Support Vector Machines (SVM) and Random Forests, were used for classifying emotions. The classification accuracies that were obtained ranged from a minimum of 66.67% for neutral emotion to a maximum of 90% for happiness. The analysis of EEG frequency bands provided different patterns for various states of emotions, with alpha waves strongly related to happiness (40%) and beta waves to anger (40%). Feature extraction times ranged from 1.25 seconds for happiness to 1.60 seconds for neutral emotion, which indicates the time efficiency of the S-Transform in processing EEG signals. These results indicate that the method using the S-Transform-based technique for emotion recognition is efficient and can be very useful for real-time applications in research and systems dealing with emotions.